Fei-Fei Li’s World Labs Unveils New World Model

World Labs, a startup led by Stanford University professor Fei-Fei Li, on Tuesday unveiled Atlas, a new world model that can generate and modify photorealistic, interactive virtual environments based on text, image and video prompts.

In a blog post, the startup said Atlas, pretrained from scratch, allows users to create a three-dimensional environment and view it from any camera position and angle they specify.

China-born Li, also known as the “AI godmother” whose research helped spark the deep learning revolution in the early 2010s, has recently pivoted her interests to world models for their potential in simulating real-world environments and approximating the physics of how objects move and interact with their surroundings. Such models could unlock breakthroughs in humanoid robots and autonomous driving, as well as creative industries like videogames and interactive content.

World Labs, which came out of stealth in 2024, faces growing competition from China, where numerous startups founded this year are racing to develop world models.

Atlas, which can generate interactive images and videos, competes directly with other world models such as Google’s Gemini Omni. Its potential market also overlaps with video generation models such as ByteDance’s Seedance 2.5 and Google’s Veo.